#1 feat: TracingChannel-powered observability

Draft — Visual comparison of TracingChannel changes on top of #3110

This PR shows the diff of the TracingChannel refactoring on top of the OTel metrics PR (redis/node-redis#3110).

What changed

Core code no longer calls OTel metrics directly. Instead it describes what's happening through diagnostics_channel:

• trace() — wraps async operations (commands, batches, connections, pool waits) via TracingChannel
• publish() — emits point events (errors, cache hits, pubsub, maintenance) via dc.channel()

OTel metrics subscribe to these channels externally. Any APM (Sentry, Datadog, custom) can do the same — zero additional surgery to core code.

Key numbers

• -312 lines removed from existing production code
• +291 lines new tracing.ts infrastructure (serves all consumers)
• Net -21 lines of production code
• Zero OTelMetrics imports in core code
• Two functions for all observability: trace() and publish()

🤖 Generated with Claude Code